1. Field of the Invention
The present invention relates to a system, apparatus, method and computer program for a wireless communication among a plurality of wireless stations, such as a communication by a wireless LAN (Local Area Network). In particular, the invention relates to such a system, apparatus, method and computer program which realize a broadband wireless transmission in home or other similar communication environments.
More specifically, this invention relates to a system, apparatus, method and computer program which enhance the transmission capacity by employing a communication where a transmitter and a receiver each having a plurality of antennas communicate with each other using space division multiplexing, that is, MIMO communication; in particular, the invention relates to such a system, apparatus, method and computer program, which are adapted to perform MIMO transmission using a singular value decomposition (SVD) of a channel information matrix each element in which represents propagation information of one of sub-channels each linking a pair of an antenna of the transmitter and an antenna of the receiver.
2. Description of Related Art
Computer networking such as LAN efficiently enables to share information and apparatus resources. Wireless LAN is attracting attention of people as a system for relieving users from the conventional wired LAN construction. In a working space such as an office, most of cables and wires can be dispensed with by employing a wireless LAN, facilitating relocation of a communication terminal such as a personal computer.
Recently, demand for wireless LAN has increased with the speed improvement and price-reduction of wireless LAN. In particular, to establish a small wireless network of a plurality of electronic devices present around people so as to enable communications thereamong, introduction of Personal Area Network (PAN) is considered. For instance, there are defined various wireless communication systems and apparatuses using respective frequency bands, e.g. 2.4 GHz and 5 GHz bands, which are permitted for use without a license from the supervisory authority.
One of the standards related to wireless networking is IEEE (the Institute of Electrical and Electronics Engineers) 802.11 (see nonpatent literature 1). IEEE 802.11 standard is further divided, depending upon the employed methods and used frequencies, into IEEE 802.11a, IEEE 802.11b . . . etc., defining respective wireless communications methods.
IEEE 802.11a standard supports a modulation method achieving a communication speed of up to 54 Mbps. However, there is a demand for a wireless standard capable of realizing a higher bit rate as the communication speed. In this situation, MIMO (Multi-Input Multi-Output) communication technology has recently attracted increased attention. This technology is for enhancing the communication speed by providing both of the transmitter and receiver with a plurality of antennas, so as to realize space division multiplexing, i.e., a plurality of sub-channels which are logically independent of one another, to increase the transmission capacity. Using the space division multiplexing, MIMO is bandwidth-efficient.
FIG. 5 schematically shows a MIMO communications system, where each of a transmitter and a receiver is equipped with a plurality of antennas. The transmitter space-time encodes N data for transmission to be multiplexed, and distributes the encoded data to M antennas of the transmitter from which the data are sent over a channel to the receiver in a multiplexed fashion. The receiver receives and space-time decodes the data received through N antennas thereof via the channel, to obtain received data. Thus, a MIMO communication is not the same as a communication by a simple transmission/reception adaptive array. In MIMO, the channel model involves an RF environment (transfer function) on the side of the transmitter, a construction (transfer function) of the channel space, and an RF environment (transfer function) on the side of the receiver. When a signal is transmitted from antennas in a multiplexed fashion, crosstalk occurs; by signal processing performed on the part of the receiver, the multiplexed signal is retrieved correctly.
A MIMO system is a communications system utilizing a characteristic of the channel. In the system, the transmitter sends out the transmitted data or signal by distributing components of the data to the plural antennas thereof (hereinafter referred to as “transmit antennas”), and the receiver obtains received data by processing the signal components received through the plural antennas thereof (hereinafter referred to as “receive antennas”). Although various applications of the MIMO transmission technology are known, one of ideal modes of MIMO is that using SVD (Singular Value Decomposition) of a propagation function, namely, SVD-MIMO system, as disclosed in Patent Document 2, for instance.
FIG. 6 schematically shows a SVD-MIMO transmission system, where a matrix of numbers, i.e., a channel information matrix H, each of whose elements represents information on each of sub-channels linking respective antenna pairs, is subjected to a singular value decomposition to obtain UDVH, and an antenna weighting coefficient matrix V on the part of the transmitter (hereinafter referred to as “transmit antenna weighting coefficient matrix V”) and an antenna weighting coefficient matrix UH″ on the part of the receiver (hereinafter referred to as “receive antenna weighting coefficient matrix UH”) are provided. Accordingly, the channel information is expressed by a diagonal matrix whose diagonal elements are square roots of respective eigenvalues λi. Thus, a signal can be transmitted in a multiplexed fashion without suffering from crosstalk at all. However, in the SVD-MIMO transmission system, it is not easy to perform the operation of the SVD in real time, and the set-up procedure such that the derived V or UH is beforehand communicated to the other part of the communication is essential.
It is possible to achieve the theoretically maximum communication capacity by the SVD-MIMO transmission system. For instance, where the transmitter and receiver respectively have two antennas, a transmission capacity of two times large at maximum can be achieved.
There will now be described the scheme of the SVD-MIMO transmission system. Where the numbers of antennas of the transmitter and receiver are M and N, respectively, transmitted signal x is represented as vector (M×1) while the received signal y is represented by vector (N×1). In this case, the channel information can be represented as a matrix H of N×M. An entry hij of the channel information matrix H represents a transfer function with respect to a sub-channel from a j-th transmit antenna to an i-th receive antenna. A vector y representing the received signal equals to a multiplication of the matrix H by the vector of the transmitted signal, plus a noise vector n, and is expressed by the following equation (1):y=Hx+n  (1)
The channel information matrix H subjected to the singular value decomposition as described above, is expressed by the following equation (2):H=UDVH  (2)
In equation (2), the transmit antenna weighting coefficient matrix V and receive antenna weighting coefficient matrix U are unitary matrices which respectively satisfy the following equations (3) and (4):UHU=I  (3)VHV=I  (4)
That is, the receive antenna weighting coefficient matrix UH is an array of normalized eigenvectors of HHH, while the transmit antenna weighting coefficient matrix V is an array of normalized eigenvectors of HHH. Further, D represents a diagonal matrix whose diagonal elements are square roots of respective eigenvalues of HHH or HHH. The size of the matrix D corresponds to the smaller one of the numbers M and N of the transmit antennas and receive antennas, that is, the matrix D is a square diagonal matrix having a rank of min (M, N).
                    D        =                  [                                                                                          λ                    1                                                                              …                                                                                                                          0                                                                    ⋮                                                                                  λ                    2                                                                                                                                                                                                                                                                                                                                                                                                                              ⋱                                                                                                                                                  0                                                                                                                                                                                                                                              λ                                          min                      ⁡                                              (                                                  M                          ,                          N                                                )                                                                                                                          ]                                    (        5        )            
In the above description related to the singular value decomposition, a case where only real numbers are involved is assumed. It is noted that in the case where imaginary numbers are also involved, even where eigenvectors of the matrices U and V, each of which is a matrix of eigenvectors, are manipulated so that the norm of each matrix is 1, that is, normalized, an infinite number of eigenvectors having respective phases, not a single eigenvector, exist. In some cases, the equation (2) can not be established depending upon the phase difference between U and V, namely, where U and V are correct but have different phases. To completely synchronize the phases, V is obtained as a matrix of eigenvectors of HHH as ordinary, while U is obtained by multiplying both terms of the equation (2) by V, as expressed by the following equation (6):HV=UDVHV=UDI=UD  (6)U=HVD−1 
The transmitter weights the components of the signal for respective sub-channels by the transmit antenna weighting coefficient matrix V, while the receiver receives the signal with weighting the components by an inverse weighting coefficient matrix UH; since each of U and V is a unitary matrix (U is represented by N×min(M, N) while V is represented by M×min(M, N)), the following expression is obtained:
                                                        y              =                                                                    U                    H                                    ⁢                  HVx                                +                                                      U                    H                                    ⁢                  n                                                                                                        =                                                                                          U                      H                                        ⁡                                          (                                              UDV                        H                                            )                                                        ⁢                  Vx                                +                                                      U                    H                                    ⁢                  n                                                                                                        =                                                                    (                                                                  U                        H                                            ⁢                      U                                        )                                    ⁢                                      D                    ⁡                                          (                                                                        V                          H                                                ⁢                        V                                            )                                                        ⁢                  x                                +                                                      U                    H                                    ⁢                  n                                                                                                        =                              IDIx                +                                                      U                    H                                    ⁢                  n                                                                                                        y              =                              Dx                +                                                      U                    H                                    ⁢                  n                                                                                        (        7        )            
The vectors y and x are not determined by the numbers of the antennas of the transmitter and the receiver, but are respectively expressed by (min (M, N)×1).
Since D is a diagonal matrix, each transmitted signal can be received without suffering from the cross talk. The amplitude of each of the sub-channels which are independent from one another is proportional to the square root of the eigenvalue λ, and thus the power of each sub-channel is proportional to λ.
As to the noise component n, since the columns of U are the eigenvectors normalized so that the norm is 1, UHn does not affect the noise power of the received signal. UHn is a vector whose size is min (M, N), which is the same size as y and x.
As described above, in the SVD-MIMO transmission, plural independent logical sub-channels free from crosstalk even occupying the same frequency band and the same time period can be obtained. This means that it is enabled to simultaneously transmit plural data using a same frequency band, improving the transmission speed.
In the SVD-MIMO system, the receiver must obtain the channel information matrix H, implement the singular value decomposition, and communicate VH as a factor of UDVH obtained as the result of the decomposition to the transmitter. In effect, the transmitter uses V and therefore V must be communicated to the transmitter.
An amount of information carried by the transmit antenna coefficient matrix V will be now considered, by taking for example IEEE 802.11a which defines one of LAN systems where the SVD-MIMO transmission is applicable, namely, OFDM (Orthogonal Frequency Division Multiplexing) of 5 GHz band.
Where each of the transmitter and receiver has three antennas, the transmit antenna weighting coefficient matrix V is a 3×3 matrix, having nine elements. In this case, when each element is a complex number represented using 10 bits, and 52 carriers are provided, a total of 9360 bits of information, i.e., 9 (the number of elements of the matrix)×2 (the real and imaginary part of a complex number)×10×52 (the number of OFDM sub-carriers), has to be fed back to the transmitter from the receiver.
The MIMO requiring such feedback is called closed-loop MIMO, while the opposite thereof is open-loop MIMO. A closed-loop SVD-MIMO system must feedback information of that much (9360 bits) to the transmitter, upon initiation of a communication. Let us assume that the information is fed back where the most reliable one in the modulation schemes provided by IEEE 802.11a, i.e., BPSK is employed as a first modulation method, the coding rate is ½, and OFDM is employed as a second modulation method. Since 1 OFDM symbol can carry only 24 bits, 390 OFDM symbols are required for the transmission of the information, making the SVD-MIMO unpractical.
As one of embodiments for realizing the above-described set-up processing in the MIMO transmission by a relatively simple mechanism, a technique called V-BLAST is known. V-BLAST is an acronym of “Vertical Bell Laboratories Layered Space Time” and refers to a technology originally developed by the now-defunct AT & T Bell Laboratories. See Patent Documents 1, for instance.
FIG. 7 schematically shows a structure of a V-BLAST communication system. A transmitter space-time encodes N data for transmission to be multiplexed, and distributes the encoded N data to M antennas (in the specific example shown in FIG. 7, two antennas) through which the multiplexed data is transmitted over a channel to a receiver. The receiver space-time decodes data received through N antennas (in the present example, three antennas) via the channel, to obtain N received data.
The difference between the V-BLAST and SVD-MIMO systems is that the transmitter in the V-BLAST does not provide the antenna weighting coefficient matrix V, but simply multiplexes a signal with respect to the transmit antennas. In other words, the feedback processing for beforehand providing the antenna weighting coefficient matrix V is all omitted. The transmitter inserts, prior to sending the multiplexed signal, training signals to be used in channel estimation by the receiver, in the multiplexed signal. For instance, the training signals for respective antennas are inserted in the signal in a time division fashion. In the example of FIG. 7, the training signals are sent included in the data packet such that a training signal Training-1 corresponding to an antenna #1 is sent following a preamble signal and a training signal Training-2 corresponding to an antenna #2 is subsequently sent, in a time division fashion.
On the part of the receiver, a channel estimator thereof performs a channel estimation using the training signals, to calculate the channel information matrix H representing information on the sub-channels linking respective antenna pairs. A first antenna weighting coefficient matrix calculator performs zero-forcing or others for each of signals corresponding to the respective transmit antennas so as to cancel unnecessary signals, i.e., signals other than that for the respective receive antennas, and obtains a receive antenna weighting coefficient matrix ZR. The transmitted signal having the highest S/N ratio among the signals retrieved after ZR is provided, is first decoded to obtain a signal x1.
Next, the decoded signal is encoded again by an encoder to generate a replica (duplicate) of the transmitted signal x1, which is canceled from the signals just received by the receive antennas. A second receive antenna weighting coefficient matrix calculator excludes the transmit antenna corresponding to the transmitted signal x1 as canceled, and again applies zero-forcing criteria to each of the other signals, to calculate a receive antenna weighting coefficient matrix ZR′. The signal x2 exhibiting the highest S/N ratio among the remaining received signals is decoded by the decoder.
In the second decoding, since the transmitted signal as decoded first is eliminated, the degree of freedom of the receive antennas is enhanced and the effect of maximal ratio combining is accordingly improved. Thereafter, all transmitted signals as multiplexed are sequentially decoded by iteration of the above-described processing.
As described above, a characteristic of the V-BLAST resides in that zero-forcing and canceling are sophisticatedly combined so that even a signal whose S/N ratio can not be made sufficiently high only by application of zero-forcing criteria can be improved in S/N ratio by taking advantage of the degree of freedom of the antennas which is provided by the canceling, and thus the accuracy of the decoding is enhanced. Thus, the V-BLAST can realize an efficient MIMO transmission system by a combination of relatively simple mechanisms.
However, since the transmitter does not perform the weighting before the data transmission, the receiver is required to implement the first decoding only by zero-forcing, without performing the canceling operation. Thus, the number of receive antennas is made larger than that of the transmit antennas so as to obtain a redundancy in degree of freedom of the receive antennas. In the example shown in FIG. 7, two transmit antennas and three receive antennas are provided.
For instance, where a bidirectional MIMO transmission system where data transmission between the transmitter and the receiver in the uplink and downlink directions are enabled is desired, the V-BLAST system as described above requires a total of 3×3 antennas at least to realize such a system, as can be seen from FIG. 8. This is because three or more antennas are essential for the receiver to assure the decoding accuracy, and since the MIMO system is bidirectional, both sides of the link should have three antennas. In a wireless LAN, for example, it is more often than not the case that providing a station of a relatively small size (mobile station) with three antennas is difficult, since there can not be spared much power source and implementation capacity for the antennas, although an access point (control station) may be able to spare more of them.